A review of temporal aggregation and systematic sampling on time-series analysis

Abstract

Purpose: This review investigates the effects of temporal aggregation and systematic sampling on time-series analysis, focusing on their influence on data accuracy, interpretability and statistical properties. The purpose of the study is to synthesise existing literature on the topic and offer insights into the trade-offs between these data reduction techniques. Design/methodology/approach: The research methodology is based on an extensive review of theoretical and empirical studies covering univariate and multivariate time series models, focusing on unit roots, ARIMA, GARCH, cointegration properties and Granger Causality. Findings: The key findings reveal that while temporal aggregation simplifies data by emphasising long-term trends, it can obscure short-term fluctuations, potentially leading to biases in analysis. Similarly, systematic sampling enhances computational efficiency but risks information loss, especially in non-stationary data, and may result in biased samples if sampling intervals coincide with data periodicity. The review highlights the complexities and trade-offs involved in applying these methods, particularly in fields like economic forecasting, climate modelling and financial analysis. Originality/value: The originality and value of this study lie in its comprehensive synthesis of the impacts of these techniques across various time series properties. It underscores the importance of context-specific applications to preserve data integrity, offering recommendations for best practices in the use of temporal aggregation and systematic sampling in time-series analysis.